Carbon footprint of diets of Norwegian households - status and potential
reductions
Aurélie Valentine Stamm
Master in Industrial Ecology
Supervisor: Edgar Hertwich, EPT
Department of Energy and Process Engineering Submission date: June 2015
Norwegian University of Science and Technology
1 I dedicate my thesis to my parents Marie -Gabrielle and Christian Stamm.
Their steady support accompanied every choice I made along my education pathway and during any moment of doubt.
2
Abstract
English
The present work aims at quantifying the current carbon footprint of diets of average Norwegian households with a holistic perspective, and assessing various ways to reduce it. The scope of the research comprises food consumed and food -related activities happening at home, including food transport , storage and preparation.
After an introduction of the context of the research, its research question, goal and scope, the report reviews existing literature. It then presents the methods used for its own analysis, and its results. The analysis is then discuss ed, and the work concludes with a brief review of its main findings and challenges.
Two models were built separately. One assesses the carbon footprint of food consumption as well as food waste, and the other makes the same assessment for food supply. Both models served to calculate their associate system’s current carbon footprint, and to build scenarios to assess reduction potentials.
The analysis finds higher emissions embodied in food consumption than in food waste and in food supply. An average Norwegian has a diet carrying emissions of 1233 kg CO2-e per year, to which wasted food adds embodied emissions of 114 kg CO2-e per year, and for which food transport, storage and preparation at home carry emissions of 203 kg CO2-e per year.
Main limitations of this analysis comes from uncertainties lying in the data.
The present work calls for further research to lower the uncertainty level, and to assess the system’s footprint on other impact categories.
3
Preface
This master thesis is carried out as par t of the Industrial Ecology master program, Department of Energy and Process Engineering, Norwegian University of Science and Technology. It models and analyzes carbon footprint of the specified system, using methods previously taught as well as independen tly developed.
The thesis was carried out under the supervision of Professor Edgar Hertwich and co - supervision of Doctor Kjartan Steen -Olsen from the Industrial Ecology program at NTNU. The goal and research question were decided in collaboration with Tro ndheim- based startup Ducky, which will use the results presented at its own ends.
Acknowledgments
I would like to warmly thank Professor Edgar Hertwich for his appreciated supervision and advices, as well as Doctor Kjartan Steen -Olsen for his precious support, patience and availability.
A special thank goes to Trent Ruder and Simon Saxegård for their welcomed proof - reading and language revision.
I would also like to thank all of my classmates from the class 2013 -2015 of Industrial Ecology, who showed support and solidarity during hard moments. The whole master program is a cherished experience in my life thank to their presence and friendship.
Last but not least, I cannot express enough gratitude to my family for supporting my moving to Norway and for u nconditionally believing in me.
4
Table of content
1 Introduction ... 8
1.1 Background ... 8
1.2 Research question, goal and scope ... 9
1.3 Thesis outline ... 10
2 Literature Review ... 11
2.1 Assessments of food consumption carbon footprint ... 11
2.1.1 Main approaches ... 11
2.1.2 Main methods ... 12
2.1.3 Main findings ... 13
2.2 Assessments of food-related activities carbon footprint ... 13
2.3 Concluding remarks ... 14
3 Methods ... 18
3.1 Carbon footprinting of current Norwegian food -related habits ... 18
3.1.1 Food consumption carbon footprinting ... 18
3.1.2 Food waste carbon footprinting ... 20
3.1.3 Food supply activities carbon footprinting ... 20
3.2 Carbon footprinting of reduction scenarios ... 24
3.2.1 Diet scenarios... 24
3.2.2 Food supply scenarios ... 32
4 Results ... 35
4.1 Assessment of the current carbon footprint of Norwegian food habits ... 35
4.1.1 Food consumption carbon footprint assessment ... 35
4.1.2 Food waste carbon footprint assessment ... 38
4.1.3 Food supply carbon footprint assessment ... 38
5
4.2 Reduction scenarios potentials ... 40
4.2.1 Diet scenarios potentials ... 40
4.2.2 Food supply scenarios potential ... 42
5 Discussion ... 44
5.1 Reliability of the results ... 44
5.1.1 Uncertainties and assumptions ... 44
5.1.2 Agreement with literature ... 47
5.2 Further implication of this work ... 48
5.2.1 Policy suggestions ... 48
5.2.2 Further research ... 51
6 Conclusion ... 53
6
List of tables
Table 1: Methodology summary from the literature reviewErreur ! Signet non défini.
Table 2: Excerpt of the multipliers tab ... 25
Table 3: Values in grams of the vegetarian scenario compared to the baseline ... 29
Table 4: Foreground to foreground matrix for scenario A ... 33
Table 5: Overview of ownerships made for each scenario ... 34
Table 6: Snapshot of the current average Norwegian diets carbon footprint results 36 Table 7: Ranking of impact reductions from every diet scenarios ... 41
Table 8: Ranking of food supply scenarios ... 43
List of Figures
Figure 1: Food supply system for all Norwegian households ... 22Figure 2: Foreground to foreground requirement matrix ... 23
Figure 3: Background to Foreground requirement matrix ... 23
Figure 4: Food supply system for scenario A ... 33
Figure 5: Overview of the annual performances of the different diet scenarios ... 41
7
List of abbreviations
CO2-e: CO2 equivalent EU: European Union FU: Functional Unit
GDP: Gross Domestic Product GHG: Greenhouse Gas
GHGE: Greenhouse Gas Emission GWP: Global Warming Potential IOA: Input-Output Analysis LCA: Life-Cycle Assessment LCI: Life-Cycle Inventory w/: with
w/o: without
WHO: World Health Organization
8
1. Introduction
1.1 Background
According to the latest Intergovernmental Panel on Climate Change (IPCC) report,
“warming of the climate system is unequivocal, as is now evident from observations of increases in global average air and ocean temperatures, widespread melting of snow and ice, and rising global average sea level” (Alley et al., 2007). Whether or not this climate change has a natural cause is a continual debate, but international experts agree to say that anthropogenic activities are “very likely” to explain its evolution since the industrial era (Alley et al., 2007). Greenhouse gases, responsible for global warming, are defined as those that “effectively absorb infrared radiation, emitted by the Earth’s surface, by the atmosphere itself due to the same gases, and by clouds”(Baede, Report, Use, Change, & Earth, 1986). The principal gas affecting the Earth’s radiative balance is carbon dioxide (CO2), principally caused by the burning of fossil fuels and land use change (Baede et al., 1986). Agriculture is the main cause of the increased emissions of the other main GHGs: methane (CH4) and nitrous oxide (N2O). Governments around the world have acknowledge d the importance of mitigating climate change; international summits are held regularly with the aim of limiting emissions that arise fr om production activities.
Hertwich and Peters (Hertwich & Peters, 2009) analyzed GHG emissions (GHGE) associated with the final consumption of goods and services on a global level.
According to their analysis, 72% of global greenhouse gas emissions are due to household consumption. Additi onally, the same authors found that out of 8 consumption categories, food consumption is responsible for 20% of the global GHGE. They also prove that food is the most important consumption category in terms of non-CO2 GHGE due to households’ demand.
In terms of environmental policy, Norway is one of the world’s pioneers and is a key influence for European Union (EU) environmental policies a ccording to the Organization for Economic Cooperation and Development (OECD 2011). In some
9 areas, Norway has adopted environmental requirements more stringent than those set by the EU.
Despite Norway’s leadership position, GHGE continue to increase and environmental problems regarding agricultural landscapes and was te generation remain of concern.
Further improvements of Norway’s environmental performance is thus not only but also desirable. Transparency and accountability are one of the keys to efficient policies. It is thus of importance for Norway to estimate its footprints in all domains and take it as a base for improvements.
Providing an assessment of the carbon equivalent emissions of Norwegian food consumption is relevant for the reasons mentioned above: the high responsibility share of households in terms of GHGE, the high contribution of food consumption, and the desirable reduction of Norwegian GHGE. The term ‘carbon footprint’ will be used to designate the CO2-e emissions arising along the life cycle stages of a product or an activity.
1.2. Research question, goal and scope
The goal of the thesis is to quantify the current carbon footprint of diets of average Norwegian households with a holistic perspective, and to assess various ways to reduce it. The research question to be answered is: What is the average carbon footprint of Norwegian diets and how can it be redu ced?
Analysis of carbon footprints associated with food consumption and household related activities fall into the scope of the thesis. This includes carbon footprint assessments of the current average food consumption for various population groups, food wastes, home food transportation, storage and preparation. Food consumed outside the home (i.e. restaurants, cantinas, hotels) falls outside the scope of this investigation.
10
1.3. Thesis outline
The first chapter will review the main existing literature releva nt to the thesis.
The second chapter presents the methods us ed for the analysis. The chapter aims at giving the reader an understanding of the methodological basis to the results.
The third chapter introduces the results for the current situation in Norw ay in all domains analyzed, as well as the quantification of reduction potential scenarios.
The fourth chapter provides a general discussion on the work done . The section gives the reader a perspective on the analysis both in terms of uncertainties that ne ed to be considered, and in terms of implications of the results that can be drawn from the analysis.
The fifth and final chapter concludes the work by briefly reviewing the main results and issues discussed.
11
2. Literature Review
Household consumption is responsible for a significant share of GHG Es, and the role of food consumption in a household’s footprint is significant (Jones & Kammen, 2011). Therefore, environmental impacts of food consumption is a relevant topic for the mitigation of climate change and researchers have looked at the question with different approaches and from different angles that will be presented hereafter.
2.1 Assessments of food consumption carbon footprint
2.1.1 Main approachesA large part of life cycle inventories (LCI) and life-cycle analyses (LCA) exist for all types of food products (Carlsson-Kanyama, Ekström, & Shanahan, 2003; Carlsson - Kanyama & González, 2009; Carlsson -Kanyama, 1998; González, Frostell, &
Carlsson-Kanyama, 2011; Wallén, Brandt, & Wenners ten, 2004). These findings on food products impacts were often associated with the conduction of impact assessments of specific meals (Carlsson-Kanyama & González, 2009; Davis, Sonesson, Baumgartner, & Nemecek, 2010; Virtanen et al., 2011) and diets (Baroni, Cenci, Tettamanti, & Berati, 2007; Gussow, 1995; Pimentel, 2003; Stehfest et al., 2009; Tukker et al., 2011; White, 2000) . In addition to assessing the environmental burdens of a given consumption pattern, some studies also assessed reduction potentials of food consumption impacts (Reijnders & Soret, 2003; Stehfest et al., 2009; Tukker et al., 2011).
The concept of “food miles” as a measure of distance that food “travels between its production and the final consumer ” (Weber & Matthews, 2008) received media and public attention in the USA and the UK. Scientists seek to scientifically test the concept (Coley, Howard, & Winter, 2009; Edwards -Jones et al., 2008; Weber &
Matthews, 2008).
Research has also been conducted to compare carbon intensities with nutrient intensities (Carlsson-Kanyama et al., 2003; Davis et al., 2010; Drewnowski et al.,
12 2015; González et al., 2011; Reijnders & Soret, 2003; Tuk ker et al., 2011; Virtanen et al., 2011; Wallén et al., 2004) , raising attention on the positive correlation between health and sustainability.
In later years, studies have mostly focused on diet analyses. Recent reviews assess the current available literature on dietary scenarios (Hallström, Carlsson -Kanyama,
& Börjesson, 2015), dietary impacts (Heller, Keoleian, & Willett, 2013) and the sustainability of dietary recommendations (Reynolds, David Buckley, Wein stein, &
Boland, 2014). Green et al. (2015) added to their review an assessment of emissions reduction potential of a diet following World Health Organization (WHO) recommendations. Instead of basing the scenarios on recommendations or hypothes es, Masset et al. (2014) worked with self-selected diets in order to propose changes that are more likely to be culturally accepted.
2.1.2 Main methods
Methodologically, LCA has been most widely used to assess food impacts (Baroni et al., 2007; Carlsson-Kanyama, 1998; Davis et al., 2010; González et al., 2011; Wallén et al., 2004), although Input-Output Analysis ( IOA) has been presented by some as the most appropriate method for footprint assessment (Duchin, 2005; Wiedmann, 2009) and used by Tukker in the context of food consumption (Tukker et al., 2011).
In 2000, Jungbluth compiled a modular LCA for calculating the impacts of t he food system in Switzerland (Jungbluth, Tietje, & Scholz, 2000) . A hybrid-LCA was used by Virtanen and compared to classic LCAs of meals (Virtanen et al., 2011). To test the food mile concept, Weber and Matthews (2008) also used a hybrid method, using IOA-LCA to assess the total freight needed from production to retail to meet food demand in the United States in 1997. Table 1 offers a detailed picture of the methodologies used for the impact assessment of food consumption.
Correlating health and environmental impa cts is often made by considering the nutrient content of different diets (Drewnowski et al., 2015; Green et al., 2015; Heller et al., 2013; Röös, Karlsson, Witthöft, & Sundberg, 2015) . Soret used mortality as an indicator of healthiness of three dietary patterns compared with their respective
13 associated GHGEs (Soret et al., 2014) and De Boer used the Body Mass Index (De Boer, Schösler, & Aiking, 2014).
2.1.3 Main findings
Most studies agree that meat and other animal products are the principal source of negative impacts from the food system in terms of global warming potential (GWP) (Baroni et al., 2007; Jungbluth et al., 2000; Stehfest et al., 2009; Wallén et al., 2004;
Weber & Matthews, 2008). Some however argue that reducing meat and dairy consumption, albeit relevant, is not enough to reach a satisfying global reduction of GHGEs (Tukker et al., 2011; Wallén et al., 2004) , calling for policy changes regarding production methods (Wallén et al., 2004). The importance of the agricultura l stage and agricultural practices is underlined by several studies (Carlsson-Kanyama &
González, 2009; Kramer, Moll, Nonhebel, & Wilting, 1999; Virtanen et al., 2011) , but the role of transport is also not to be neglected (Coley et al., 2009; Edwards -Jones et al., 2008; Jungbluth et al., 2000) .
Finally, aside from carbon emissions and global warming potential, severa l authors call for research on the bigger picture of food impacts, notably depletion of fish stocks (Gussow, 1995; Tukker et al., 2011) , land use (Tukker et al., 2011), water consumption and the ethics behind the production and distribution of food (Baroni et al., 2007).
2.4. Assessments of food-related activities carbon footprint
Fewer studies have been done to assess the environmental impacts arising from home - made food preparation. Carlsson -Kanyama et al. (2003) included it in the system boundaries for their LCA of food products, but did not present the specific emissions arising from that stage. In 2005 , Sonesson et al. performed a detailed analysis of the consumer phase including home transport, cooking, storing and wastage and argue for the inclusion of those least investigated activities in food system analyses (Sonesson, Anteson, Davis, & Sjödén, 2005) . A similar study followed five years later, when
14 Kauppinen and colleagues presented a broken-down analysis of the food-preparation activities for the particular case of Finnish households (Kauppinen, Pesonen, Katajajuuri, & Kurppa, 2010). Sonesson and colleagues compared the environmental impacts between homemade and industry-made meals (Sonesson, Mattsson, Nybrant,
& Ohlsson, 2005); a similar comparison followed in 2014 (Schmidt Rivera, Espinoza Orias, & Azapagic, 2014) .
2.5. Concluding remarks
It appears clear that, although food product impacts have been extensively assessed, there is a need for extending the system boundaries in the studies of food. Compiling food products into specific diets, and then using those diets to build up scenarios have been done several times. Furthermore, the definition of diets has been limited to the food intake; food preparation was systematically assessed in separate studies, or not specifically underlined.
This work aims at adding to the research an assessment of the Norwegian -specific diet and its carbon footprint with an integration of the meal supply processes.
15
Table 1: Methodology summary form the lit erature revi ew
Data Allocation /
disaggregation
Assessment method
System boundaries
FU Software Impact Ass. Time and
Space
Other
Tukker et al. (2011)
E3IOT
environmentally extended input output database;
Eurostat; Concise European Food Consumption Database; FAO (Food Balance Sheet).
50 food groups. Food processing allocated to each food group, but treated as separate category for households (assume similar impact in all diets).
EIOA. Country cluster for diets. Each cluster based on their ratio vegetal/animal.
3 scenarios, each illustrating diff ratios, hence small dietary shifts.
Does not include water use and land use.
Ratio
vegetal/animal per diet
E3IOT &
CAPRI (partial equilibrium model for rebound effect).
Recommends ReCiPe for impact assessment.
Climate change, ODP, AP, EP, Human toxicity, Photochemical Oxidant Formation, Ecotoxicity, Abiotic Resource Depletion.
Europe, 2003
Takes into account the rebound effect of 1 and 2 order.
González et al.
(2011)
Statistics; IPCC (2006); literature;
IEA (2009).
Not displayed. LCI of 84 food items.
From cradle- to-
Gothenburg Port
1 kg of food product delivered to Gothenburg port.
FU of meat:
bon-free carcass.
FU of cereals and beans: 1 kg od dry grain at the port (thus excluding packaging).
Not displayed.
Energy use and greenhouse gas emissions.
Sweden, from 2003 to 2011
Takes into account the protein efficiencies.
16
Virtanen et al.
(2011)
EIO-LCA: Finnish economic IO table.
LCA: literature;
Finnish farming database;
Industries;
national statistics;
personal communications;
EcoInvent.
EIO-LCA:
disaggregation with help of MFA. Sub- sectors have only 1 sub- prod LCA: not displayed.
EIO-LCA of Finnish food chain and
"Lunch plate approach"
(process-based LCA of 30 lunch plates).
LCA: From agricultural production to consumer.
EIO-LCA:
Finnish Food chain, 2005.
LCA: one lunch plate (repeated for 30
different), based on Finnish standards for omnivorous, vegetarian and vegan.
Ecoinvent ENVIMAT- model
CO2, CH4, N2O, PFC, NH3
Finland, 2005.
Equal amount of energy for all lunch plate, equal share of proteins, carbohydrates, vegetables etc.
Davis et al. (2010)
Survey, reports and Ecoinvent.
Economic. LCA of 4 meals composed of different amounts of soybeans and peas both in terms of direct and indirect consumption.
From Cradle- to-grave.
Production of food, fertilizers and fuels.
Packaging (production and waste).
Electricity
&heat.
Transport.
Sewage treatment.
A meal served at the household (with the same nutritional value).
Not displayed.
Used Ecoinvent database.
Use of renewables, of non-
renewables, GWP, Photo oxidant formation potential, NOx level,
stratospheric ozone DP, EP, AP.
Spain &
Sweden, 2005 -2006
Weber &
Matthew (2008)
US input-output accounts for total economy-wide household expenditure on food and food availability statistics in the US.
By caloric ratios of the primary food group.
IO-LCA. Total freight distance from production to retail to meet food demand in 1997.
Economy-wide and per-capita data were normalized to the common unit of household.
Not displayed.
Results presented in CO2 equivalent.
USA 1997 Only assesses transport- related emissions
17
Baroni et al. (2007)
Average Italian food
consumption.
Textbooks and scientific papers.
Subsume food products into categories, with weighting depending on consumption data.
LCAs of 3 weekly diets:
omnivorous, vegetarian, vegan;
combined with 2 production method:
conventional and organic.
Not displayed. A weekly diet with a certain energetic and nutrient content.
SimaPro5, Eco indicator 99 for assessment phase.
Damages to human health, damages to ecosystems quality, damages to resources.
Analyzed with the 3 cultural approaches.
Italy Results are given in points.
Results also calculated for the
conventional Italian diet for comparison.
Wallén et al. (2004)
National statistics, published studies, manufacturers.
Not displayed. LCA of food groups and produces.
Processes included in the cultivation and
distribution of food.
Energy use and emission of GHG for food production, processing, and distribution needed for food consumption per capita in Sweden in 1 year.
Not displayed.
Energy use and CO2
equivalents.
Sweden, 1999.
Compare the consumption level of 1999 of different food groups with a consumption level suggested by another study and presented as a sustainable diet.
Jungbluth et al.
(2000)
Review of 150 studies
investigating life cycle of food products; Swiss agricultural production inventory; Swiss consumption patterns.
5 modules:
type of agricultural practice, origin, packaging material, type of preservation and
consumption.
Modular LCA. From cradle- to-grave:
From agricultural production to the end-of-life management.
1 kg of purchased product.
Eco-indicator 95,
Ecoinvent.
All impact categories.
Switzerland, 1999
Also took into account different agricultural production methods.
Carlsson- Kanyama
(1998)
Previous studies Not displayed. LCA of carrots, tomatoes, potatoes, pork, rice and dry peas
All processes from the production chain prior to the
consumer's purchase of food.
1 kg of the produce sold by Swedish retailers during the early mid- 1990s.
Not displayed.
Results presented in CO2 equivalents with a 20 year time
perspective.
Sweden mid-1990s.
Results for GHGE and energy consumption.
18
3. Methods
3.1 Carbon footprinting of current Norwegian food-related habits
3.1.1 Food consumption carbon footprintingThe first step of the work was to assess the carbon footprint of current average diets in Norway. To do so, data were needed on two sides: the average food consumption per food item and the life-cycle carbon intensities per food item .
3.1.1.1. Food consumption data
Food consumption data was taken from the Norwegian dietary survey “Norkost”
(Totland, Melnæs, Lundberg-Hallen, Helland-kigen, & Lund-blix, 2012), which discloses the average food intake of Norwegians aged 18-70, broken down by gender, age group (‘18-29’,’30-39’,’40-49’,50-59’,’60-70’), as well as households type (‘family with kids’, ‘family without kids’, ‘single’). Food products are grouped in 16 categories while beverages are grouped in 7 categories. As water is an essential human nutrient, its carbon footprint was excluded from this study. Thus 22 food and beverages categories are retained for the pre sent work.
The greatest level of detail was desirable in order to provide an analysis as precise as possible. For this reason, national statistics on food and beverages consumed per person and per year were used to complement the Norkost survey (SSB Table 10249).
This was the case for ‘fresh/frozen vegetables’ to which 5 vegetables types were added, for ‘fresh fruit and berries’ to which 7 types of fruits were added, for ‘pure red meat’ with the addition of 3 meat types, ‘sugar, honey, sweet spreads’ with 2 items added and ‘chocolate, candy’ with also 2 items added. This gives a total of 19 food items taken from SSB to complement Norkost food categories.
The food items were grouped in 16 categories, some of which had up to 5 subcategories. In total, 85 foo d items were specified in the present analysis.
Beverages were grouped into 6 categories, including a total of 15 products.
19 There are thus a total of 22 food and beverage categories and 100 broken -down items.
Three beverages categories (‘beer, ‘wine’ and ‘liquor’) and two food categories (‘juice and mash’ and ‘eggs’) are not further broken-down.
It was systematically ensured that the consumption intake of a food/beverage category in grams was equal to the sum of the intake of its components.
3.1.1.2. Carbon intensities data
Carbon intensities were collected for each of the food and beverage items presented above. Existing literature was the main source of data, complemented by Environmental Product Declarations (EPDs) and own assumptions where information was lacking. Carbon intensities at the aggregate level of each category was calculated as the weighted average of its components relative to the consumption values.
Seventy-three food products intensities were collected from existing literature, among which seven from EPDs. Assumptions based on similar food products helped assign carbon intensities for 27 food items, among which 4 were averages of other food items and 23 were assumed equal as another food item . Three food items were assessed through informed guesses.
Carbon intensities are given in CO2-equivalents per food volume, noted CO2-e, as carbon dioxide is the reference gas to measure greenhouse effect of other gases
3.1.1.3. Calculations
Carbon footprints of current diets were assessed by multiplying the two types o f information previously collected (consumed amounts and carbon intensity) of each food product and category. Results are given both as a broken -down table comprising all the food and beverage types and as a table of aggregates only including the 22 food and beverage categories.
20 3.1.2 Food waste carbon footprinting
Impacts from food wastes were calculated using the model built for the carbon footprint of diets. Information on food waste in Norway is scarce but European food waste levels were found in an FAO report on global food losses (Gustavsson, Cederberg, & Sonesson, 2011). The report gives percentages in weight of food wasted at the consumer stage and per food category. It presents seven categories (‘cereals’,
‘roots & tubers’, ‘oilseeds & pulses’, ‘fruits & vegetables’, ‘meat’, ‘fish & seafood’
and finally ‘milk’). To assess food waste, these categories were matched to the ones used in the present work. For any of the 22 categories for which the waste percentage could not be matched, conservative assumptions were made and the lowest percentage given by the report – that is 4% for ‘oilseeds & pulses’ - was applied. This is the case for ‘sugar & sweets’, ‘miscellaneous ’, ‘soft drinks and soda’, ‘eggs’, ‘cakes’ and
‘grain products’. Goods with long shelf-lives such as tea, coffee and alcohols were assumed to have a waste degree of 1%.
Based on the average Norwegian diet gathered for the ass essment of its carbon footprint, waste percentages were used to increase their respective food category consumption level. Food wastes are thus allocated to an increase in food consumption, acting as an unused demand for food. The artificially increased f ood consumption was then multiplied by the carbon intensities (in kg CO2-e/ kg product) presented in the second paragraph of this section, thus giving results representing a diet accounting for food wastes.
3.1.3 Food supply activities carbon footprinting
Apart from the emissions arising from the food items themselves, it was interesting to investigate the emissions coming from food -related activities that are food transportation, storage and preparation from a household.
In order to assess the carbon footprint of such activities a simple life cycle analysis (LCA) was conducted. Life cycle analysis is a tool to assess environmental impacts
21 arising all along a product or an activity’s life cycle in a holistic manner, i.e. from its production phase, including extraction of materials, to its end-of-life. The LCA conducted here only assessed the use phase of the selected processes. This was a conscious choice based on the goal and scope of the present work that is to evaluate the carbon footprint of diets and its red uction potential at a household level. The composition of a diet, as well as the means used to transport, store and prepare the food are variables upon which a household can make conscious choices. However, an average household cannot influence the product ion and end-of-life processes of appliances and food products that happen at an industrial and agricultural level. As such, it was deemed more relevant to focus the analysis on variables of a diet that can be influenced by personal choices.
LCAs are built with matrices representing the foreground processes requirements (Af f), the background processes requirements (Ab b) and the upstream inputs of background processes to foreground processes (Ab f) (Strømman 2010).
The system boundaries were drawn to cover th e most relevant processes in the Norwegian context. Ten processes composed the foreground: gas stove/oven, electric stove/oven, microwave, fridge without a freezer (‘fridge w/o freezer’), fridge with a freezer (‘fridge w/ freezer’), separate freezer (‘free zer’), electric car, conventional car, food preparation, food storage, food transport and finally food supply. Three processes composed the background: gas , electricity and fuel. Figure 1 illustrates the system thus composed.
22
Figure 1: Food supply system for all Norw egian households
The functional unit (FU) designates the external demand placed upon the system and is here defined as one year of food supply for all Norwegian households.
Data were taken from national statistics an d from a SINTEF report (Hanne, Rosenberg, & Feilberg, 2010) . Background processes were selec ted from the ecoinvent 22 database (ecoinvent 2015). The Arda software (Majeau-Bettez &
Strømman 2014) was chosen to build the system an d compute the results. Figures 2 and 3 show respectively the foreground requirement matrix (Aff) with the associated final demand (yf) and the requirements placed by the foreground to the background matrix (Ab f) as they will be read by Arda.
23
Figure 2: Foreg round to foreground requirement matrix
Figure 3: Background to Fo reground requirement matri x
Ecoinvent processes are the following:
2098: Natural gas, high pressure/ at consumer/ DK/MJ.
1124: Electricity, low voltage/ at grid/ NO/kWh
2764: Operation, passenger car/ electric/ LiMn2O4 / CH/ km
2755: Operation, passenger car/ RER / km.
3.1.3.2. Sensitivity analysis
From the results of the baseline analysis, it was obvious that the most influential process was the use of conventional cars to transport food from the groceries. The paper used for data on car trips conducted a survey i n Sweden in 2005, which specifically assessed trips with the sole purpose of grocery shopping. However, accounting so means leaving out combined trips, such as going back home after work and taking a detour to shop for food on one’s way. The distance drive n to transport food from the grocery shop to the house is then substantially uncertain, in addition to
24 being the most significant process in the food supply system. This is why it was of particular importance to assess the system’s sensitivity to this para meter.
To do so the distance driven was increased by 10% in the background to foreground requirement matrix (Ab f) while leaving all other parameters equal to the baseline. A new analysis was launched with Arda, and total impacts between the new system and the baseline were compared.
Another variable that was judged potentially significant is the choice of the Ecoinvent process for electricity use. This background process is determinant for 5 of the foreground processes and it is unsure whether the one chos en is the best representative of the actual electricity use and its associated carbon intensity. To test the sensitivit y of the system to this background process, another analysis was launched replacing
“electricity, low voltage/ at grid/ NO/kWh” by “elect ricity mix/ NO/ kWh” in Arda.
A last potentially significant uncertainty is the ownership of gas stoves and ovens.
The value used in the baseline is an assumption made from the ownership value of electric stoves and ovens found in the SINTEF report. There , electric oven/stoves are said to have an ownership of 96%; the remaining 4% was assumed to be ownership of gas stoves and ovens, thus assuming that any household have either one of the two systems, leaving no ownership to for instance wood -fired stoves. To assess the error margin that this assumption could bring to the results, a new analysis with an ownership of 1% gas stoves and 99% electric stoves was launched.
3.2. Carbon footprinting of reduction scenarios
3.2.1. Diet scenariosReduction potentials were compu ted through the model made to assess the current carbon footprint of diets. This assessment was made on a weight basis, though to ensure building realistic scenarios, all values in kilograms from the original model were converted to kilocalories. The total calories intake was kept constant for all scenarios - but the first one (see section 3.3.1 paragraph “Scenario 1: 2300 kcal” ) - in order for them to be comparable.
25 In order to do the conversion, calories intensities in kcal/100g food were gathered from Matvaretabellen for each food products (Matvaretabellen 2014). Calorie intensities of the food groups are the simple average calories of their components.
Table 1 gives an insight of the three different types of multipliers used for this work.
It presents the multipliers of the first 21 products and 5 food categories.
Table 2: Excerpt of the mult ipliers tab
The analysis of current diets was run a second time based on kcal to ensure consistency of the model. Results found in the two units however differ by emissions
26 coming from tea consumption, as tea is assumed to have no calories in Matvaretabellen.
For practical reasons all scenarios were built at the aggregate level of food products (22 categories) and for the population groups o f men, women as well as for the average of the two. The baseline scenario is the current consumption of these selected population groups.
Now will be introduced each scenario and its specific calculations.
3.2.1.1. Scenario 1: 2300 kcal
A first observation from the baseline is that average men and women’s diet reach kcal intakes that exceeds the recommended calories intake (USDA, 2010). As such it was relevant to evaluate the carbon benefits of eating the recommended amount of calories without necessarily changing the structure of one’s diet. To do so , the baseline diet was rescaled to reach a 2600 kcal for men and 2000 kcal for women, thus giving a 2300 kcal in average. Each food category keeps the same contribution percentage in the diet as in the baseline. Put i n simple words, people don’t change their habits but only resize their portions.
Rescaling was achieved as follow:
kcal1n = (kcal0n * 2300)/kcal0t ot
Where
- Kcal1n expresses the amount of kcal ingested from the nt h category after rescaling (scenario 1)
- Kcal0n expresses the amount of kcal ingested from the nt h category in the baseline (scenario 0).
- Kcal0t ot expresses the total kcal ingested in the baseline.
The new intakes were then multiplied with the CO2 intensities of each category as well as the grams intensity.
27 3.2.1.2. Scenario 2: Pescetarianism
This scenario analyzes a diet from which meat consumption is excluded. That is to say that pescetarians eat animal products such as fish, eggs and dairies but stay away from meat and meat products. To make the assessm ent, the intake percentage of ‘meat and meat products’ was set to 0. To keep the kcal constant, the original contribution of meat was redistributed to vegetables, fish and eggs. This was done by raising the contribution percentage of these categories by th e original contribution percentage of
‘meat and meat products’ divided by 3.
Redistributing the calories from meat was done as follow:
kcal2n = kcal0t ot * (P0n + (P08/3)) Where
- Kcal2n is the kcal ingested in scenario 2 coming from the nt h category.
- P0n is the original contribution percentage of the n category.
- P08 is the original contribution percentage of the 8t h category, namely
‘meat and meat products’.
- Kcal0t ot expresses the total kcal ingested in the baseline.
New intakes where then each multiplied by CO2 intensities and grams intensities.
3.2.1.3. Scenario 3: Vegetarianism
This scenario simulates a diet that excludes both meat and fish products. The same reasoning as for building the second scenario was applied.
The contribution percentages of fish and me at products in the baseline diet were summed up to give the proportion of calories intake to be redistributed in the scenario.
It was chosen to redistribute the lost calories by increased intakes of ‘grain products’,
‘vegetables’, ‘eggs’, ‘milk and yoghurt’, and ‘cream and cream products’.
It is assumed that a vegetarian diet is based on a high intake from vegetables, as indicated in the dietary recommendations from USDA (2010). As meat and fish products make up together for 15% of the calories intake in the baseline, each of the
28 category mentioned above was augmented by 15
7 percentage points, except for the
‘vegetables’ category which was augmented by (15
7 × 3) percentage points. As USDA recommendations are given on a weight basis the values in grams allow to verify the validity of the scenario. Table 2 shows the intakes in grams of each category for the baseline and the vegetarianism scenario.
Meat and fish categories were both set to 0. The redistribution in the selected categories mentioned above was done as follow:
Kcal3n = kcal0t ot * (P0n +15
3)
Where
- Kcal3n is the kcal ingested in scenario 3 coming from the nt h category.
- P0n is the original contribution percentage of the n category.
- Kcal0t ot expresses the total kcal ingested in the baseline.
Redistribution in the vegetables category was done as follow:
Kcal35 = kcal0t ot * (P05 +(15
3 × 3))
Where
- Kcal35 is the kcal ingested in scenario 3 coming from the 5t h category, namely ‘vegetables’.
29
Table 3: Values in grams of the veg etarian scenario compared to the basel ine
30 3.2.1.4.Scenario 4: Veganism
This scenario tests the impact of excluding any animal product, namely meat, fish, eggs, dairies and honey. Honey was here not excluded for the reason that it is aggregated in the ‘sugar and sweet’ category. However this is a minor shortcoming because sugar products only makes up for 0.08% of the food consumed by an average Norwegian, which corresponds to 0.1% of the total carbon footprint. Butter was also not possible to exclude because it is comprised with margarine in the category “butter, margarine and oil”. The broken-down baseline diet is not disaggregated enough to be able to isolate butter from margarine. It would by consequence be not realistic to set
‘butter and margarine’ to 0 since vegans need an intake of fat, which comes for a big part from oil and margarine.
The categories that were removed from this scenario are thus ‘meat and meat products’, ‘fish and fish products’, ‘eggs’, ‘milk and yoghurt’, ‘cream and cream products’ and finally ‘cheese’. Animal products make up together for 31% of the baseline kcal intake. The compensating foods in this scenario are ‘grain products’,
‘cakes’, ‘potatoes’, ‘vegetables’, ‘fruit and berries’ and ‘juice and mash’. Here too a special emphasis was put on the ‘vegetables’ category. This diet has similar calori es intake proportion from ‘bread’, ‘vegetables’ and ‘fruit and berries’.
The redistribution to ‘grain products’, ‘cakes’, ‘potatoes’, ‘fruit and berries’ and
‘juice and mash’ was done as follow:
Kcal4n = kcal0t ot * (P0n +31
8) Where
- Kcal4n is the kcal ingested in scenario 4 coming from the nt h category.
- P0n is the original contribution percentage of the n category.
- Kcal0t ot expresses the total kcal ingested in the baseline.
The redistribution to ‘vegetables’ was done as follow:
Kcal45 = kcal0t ot * (P05 +(31
8 × 3))
31 Where
- Kcal45 is the kcal ingested in scenario 4 coming from the 5t h category, namely ‘vegetables’.
3.2.1.5. Scenario 5: Vegetarian dinners
Here is tested a diet that only exclude s meat and fish at dinners but allows those products for other meals of the day. It was assumed that dinner accounts for 60% of the kcal intake in a day. To test the impacts of vegetarian dinners, both meat and fish categories were thus decreased by 60%.
To compensate for the calories lost, two steps were necessary. First th e percentage loss in kcal brought by the decrease in meat and fish intakes was calculated. This corresponds to a 9% loss of calories. To spread this 9% of missing intake the same method as for the second, third and fourth scenarios was used, that is to increase the contribution percentage of compensating categories by the percentage points lost divided by the number of compensating categories. Here an equal increase of intake from ‘bread’, ‘grain products’, ‘potatoes’ and ‘vegetables’ was assumed. Assuming so implies that a person shifting from the current average Norwegian diet to such a diet will not fundamentally change his/her cooking habits and will rather increase the portions of food items that usually accompany meat at dinner, namely carbohydrates , such as bread. Redistribution was done as follow:
Kcal5n = kcal0t ot * (P0n +
(
𝑃0𝑡𝑜𝑡−𝑃5𝑡𝑜𝑡.𝑜𝑟𝑔 4 )) Where
- Kcal5n expresses the amount of kcal ingested from the nt h category after rescaling (scenario 5)
- Kcal0t ot expresses the total kcal in gested in the baseline - P0n is the original contribution percentage of the n category
- P0t ot is the total percentage intake of the baseline coming from all food and beverages categories (thus equal to 100%).
32 - P5t ot . org is the total percentage intake of the 5t h scenario coming from all food and beverages categories after decreasing meat and fish intake from dinners but keeping a total kcal constant form the baseline (thus equal to 100% - 91%).
3.2.1.6. Scenario 6: Decreased dairy intake
As ‘milk and yoghurt’ and ‘cheese’ are respectively the second and third most impacting food categories in the current diet, this scenario tests the benefits of decreasing calories intake of dairies by 50%.
The same method as for the 5t h scenario - vegetarian dinners - was applied. The decreased categories are ‘milk and yoghurt’, ‘cream and cream products’ and ‘cheese’
by 50%. Compensating calories come from ‘bread’, ‘vegetables’, ‘eggs’ and ‘butter, margarine, oil’, which were all equally increased. As mentioned for scenario 4, butter as a product is not possible to disaggregate from margarine. Yet , it is realistic to assume that a decrease in cream products will likely be replaced by an increase in other fat products, such as margarine or oil. Cream and cream products, as well as cheese, serve approximately the same purpose in a diet than products comprised in
‘butter, margarine and oil’ category, such as preparing a sauce, accompanying a meat or fish or spreading on a slice of bread.
All scenarios were multiplied by 30.5 and by 365 to show results respectively per month and per year.
3.2.2. Food supply scenarios
In order to test the best combination of appliances and car choices , a series of seven scenarios were prepared. The functional unit was changed to ‘a year of food supply for one Norwegian household’. The foreground requirements were changed from the baseline in order to meet the needs of a single household. To present the method more explicitly, the first scenario will be taken as an example: it represent s a household
33 owning 1 gas stove, 1 microwave, 1 fridge with freezer, 1 extra freezer and 1 conventional car. Table 3 shows the modified foreground system.
Table 4: Foreground to foreground matrix for scenario A
The corresponding system is displays in figure 4 .
Figure 4: Food supply system for scenario A
Such modifications to the foreground allowed for the construction of six other scenarios. Each corresponding foreground matrices may be found in Appendices A - G.
A short summary of scenarios is shown in table 4.
34
Table 5: Over view of own er ships made for each scenari o
Gas
stove/oven
Electric stove/oven
Microwave Fridge w/o freezer
Fridge w/
freezer
Separate freezer
Electric car
Conventional car
A x x x x x
B x x x x x
C x x x x x
D x x x x
E x x x x x
F x x x x x
G x x x x
With 8 variables with which to compose scenarios, there are 8! = 40 320 different combinations possible. The seven selected here are believed to be the most inter esting for the purpose of this work and/or closer to reality in the Norwegian context.
35
4. Results
The results obtained through the methods aim at answeri ng the main research question: What is the average carbon footprint of Norwegian diets and how can it be reduced?
4.1. Assessment of the current carbon footprint of Norwegian food habits
4.1.1 Food consumption carbon footprint assessment 4.1.1.1. Overall results
According to the present analysis, an average Norwegian diet leads to emissions of 1233 kg CO2-e/ year. This is equivalent to 3.38 kg CO2-e/ day or 102.7 kg CO2-e/
month.
On average, women eat less than men, which consequently leads to a food consumption carbon footprint 1.3 times lower compared to the one of men. An average man’s food consumption leads to emissions of 1.4 tCO2-e/ year while a woman ’s food consumption carry embodied emissions of 1.0 tCO2-e/ year. On average, a woman’s diet thus leads to 26% less emissions than a man ’s diet, equal to 367 kg CO2-e/ year of saved emissions.
Accounting for the Norwegian population in 2015 (SSB Table 05810), the Norwegian male population’ food consumption leads to emissions of 3.7 Mt CO2-e/ year while the Norwegian female population ’ food consumption leads to emis sions of 2.7 Mt CO2-e per year. In total, 6.4 Mt CO2-e per year are emitted due to Norwegian food consumption.
Table 5 provides a snapshot of the results in g CO2-e arising from the food consumption of some 30 food products and 6 categories.
36
Table 6: Snapshot of the current a verage Norw egian diets carbon footprint result s
37 4.1.1.2. Most impacting food groups
The two most carbon intensive food gro ups are first ‘cheese’ with an intensity of 9 kg CO2-e / kg of cheese and then ‘meat and meat products’ with an intensity of 6 kg CO2- e/kg of meat.
The two least carbon intensive food groups are first ‘potatoes’ with an intensity of 0.3 kg CO2-e/kg of potatoes and ‘fruit and berries’ with an intensity of 0.4 kg CO2- e/kg of product
In beverages, ‘wine’ and ‘liquor’ are actually as intensive as meat as they both present a carbon intensity of 6 kg CO2-e/ kg of product.
However in a diet the most and least c arbon intensive food categories might not be the most and least impacting consumption categories as the overall footprint also depends on the amount of food consumed in each category. This way, for the average Norwegian diet, the two most impacting food gr oups are first ‘meat and meat products’ with 331 kg CO2-e/ year (27% of total impacts) and second ‘milk and yoghurt’ with 175 kg CO2-e/ year (14% of total impacts). Following these come
‘cheese’ with 152 kg CO2-e/ year and ‘wine’ with 98 kg CO2-e/ year.
4.1.1.3. Most impacting social groups
Men between 18 and 29 years old and men between 30 and 39 years old are the two most impacting social group. They are however the two social groups consuming the most calories, so such a ranking in terms of carbon footprint is n ot surprising.
The least impacting social groups are women between 60 and 70 years old and women without kids. Women in their 60’s are also the group consuming the least calories, but this does not hold for women without kids, as women between 18 and 29 y ears old are the second least calorie consuming group. The difference can then be explained by eating habits: women without kids and women in their 60’s are the two groups eating the less calories form ‘meat and meat products’ and ‘cheese’, which are the two most intensive food groups.
38 4.1.2 Food waste carbon footprint assessment
After including food wastes in the diet , 30 kg of food per year and 6.5 kg of beverages were added to the Norwegian average diet. The se correspond to an additional impact of 114 kg CO2-e/ year/ person. Given Norway’s population in January 2015 , food wastes amount to 190 kt of food and beverages per year, resulting in 588 kt CO2-e indirectly emitted.
This increase in amount of food does not result in a linear increase in CO2 release.
Although food wastage increases the total food volume by 4%, it increases the related carbon emissions by 9%. This is explained by the fact that meat products, fish products and dairy products are not only the most carbon intensive food categories, but also some of the most wasted ones. Cereals, such as bread, is the single most wasted product category according to the percentages given by FAO. However, this is not of such impact in terms of carbon release as bread has a low carbon intensity (0.99 kg CO2-e / kg product) compared to meat (6.13 kg CO2-e / kg product) or cheese (9.5 kg CO2-e / kg product).
4.1.3 Food supply carbon footprint assessment
Norwegian households’ food supply activities lead each year to emissions of 1.07 MtCO2-e. In comparison, Norwegian ho usehold’s food consumption lead to 6.4 Mt CO2-e as presented in section 4.1.1.
The most impacting process of the system is the use of conventional cars to shop for food, emitting 954 kt CO2-e. The second most impacting process is the use of separate freezers with 48 kt CO2-e emitted each year. The first process is thus about 19 times more impacting than the second one. The use of electric stoves and ovens comes third with emissions of 28 kt CO2-e.
The least impacting process is the use of microwaves, which emits 318 t CO2-e each year. The second least impacting process is the use of electric cars to shop for food, emitting 586 t CO2-e. A big difference is observed between the least impacting process and the second least impacting one, as the use of electric cars by all Norwegian
39 households has less than half of the direct emissions than the use of microwaves. The use of gas stoves and ovens is the third least impacting process with emissions of 2 kt CO2-e per year.
The most impacting process of the system e mits 2 995 times more than the least impacting one.
Both types of fridge are the processes with median emissions. A fridge with an integrated freezer is unsurprisingly more impacting ( 26 kt CO2-e) than a fridge without (16 kt CO2-e) since it has a higher ownership share and a higher electricity consumption.
From the Food Supply system, an average household carries embodied emissions of 447 kg CO2-e / year, which also amounts to 203 kg CO2-e/ pers / year.
4.1.3.2. Sensitivity analysis
The first parameter tested was the distance driven by car. Results showed a high response of the system to this parameter. After increasing the distance by 10% , overall impacts increased by 8.8 %, showing substantially strong correlation between this parameter and the system’s performance.
The second parameter tested was the choice of background process in ecoinvent 22 to represent electricity use. Changing this parameter from ‘Norwegian electricity/ low voltage/ at grid’ to the ‘Nordic electricity mix’ decreased overall impacts by 3%. The uncertainty behind the ecoinvent process is thus not particularly worrying for the reliability of the results.
The ownership percentage of gas stoves and ovens was thirdly tested. A 0.05%
reduction in overall impacts is observed after lowering the ow nership percentage by three percentage points, from 4 to 1%. Here too the system’s environmental performance seems to be little dependent on this parameter.
40
4.2. Reduction scenarios potentials
4.2.1. Diet scenarios potentialsThe most effective scenario is scenari o 4 – veganism. Compared to the baseline, it allows for a reduction of impacts of 39%. This diet leads to emission reduction of 1.3 kg CO2-e per day, which is equal to 40 kg CO2-e per month and 479 kg CO2-e per year.
The least effective scenario is scenari o 6 – decreased dairy intake. It leads to a reduction of impacts of 9%, which amounts to a reduction of 0.3 kg CO2-e per year, 9 kg CO2-e per month and 110 kg CO2-e per year.
Scenarios 2 and 3 – pescetarianism and vegetarianism – show very similar results.
Both lead to impact reductions of 17%. Pescetarianism is a slightly more efficient scenario than vegetarianism, as the former leads to emissions of 1010 kg CO2-e a year while the latter leads to emissions of 1011 kg CO2-e a year.
Scenarios 1 and 5 – 2300 kcal and meat-free dinners – have comparable results to scenarios 2 and 3. Scenario 1 gives a 16% impact reduction while scenario 5 gives a 15% impact reduction.
Scenarios 1, 2, 3 and 5 – respectively 2300 kcal, pescetarianism, vegetarianism and vegetarian dinners – all show impact reductions comprised between 15 and 17%.
Figure 5 displays each scenario’s carbon footprint and thus helps visualizing the different associated reduction potentials.
